Importing all datasets with the molecules

After the virtual screening with each QSAR model, let’s concatenate all the outputs to analyze the molecules and ranking them according to different scores

Importing output from virtual screening of each QSAR model

Index of the models

indices_2 <- read.csv("indice_2.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
indices_2 <- cbind(indices_2, c(1:310))
colnames(indices_2) <- c("Label", "Index")

indices_3 <- read.csv("indice_3.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
indices_3 <- cbind(indices_3, c(1:465))
colnames(indices_3) <- c("Label", "Index")

QSAR Results

## LASSO VS
library(knitr)
library(dplyr)
## 
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
LASSO_2 <- read.csv("VS_LASSO-2.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
LASSO_2 <- left_join(LASSO_2, indices_2, by = "Index")

LASSO_3 <- read.csv("VS_LASSO-3.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
LASSO_3 <- left_join(LASSO_3, indices_3, by = "Index")

## LASSO-MLR VS
L_MLR_2 <- read.csv("VS_MLRLASSO-2.csv", stringsAsFactors = FALSE, header = TRUE, 
    sep = ",")
L_MLR_2 <- left_join(L_MLR_2, indices_2, by = "Index")

L_MLR_3 <- read.csv("VS_MLRLASSO-3.csv", stringsAsFactors = FALSE, header = TRUE, 
    sep = ",")
L_MLR_3 <- left_join(L_MLR_3, indices_3, by = "Index")

## GA-MLR VS
GA_MLR_2 <- read.csv("VS_GAMLR-2.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
GA_MLR_2 <- left_join(GA_MLR_2, indices_2, by = "Index")

GA_MLR_3 <- read.csv("VS_GAMLR-3.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
GA_MLR_3 <- left_join(GA_MLR_3, indices_3, by = "Index")

## PLS VS
PLS_2 <- read.csv("VS_PLS-2.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
PLS_2 <- left_join(PLS_2, indices_2, by = "Index")

PLS_3 <- read.csv("VS_PLS-3.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
PLS_3 <- left_join(PLS_3, indices_3, by = "Index")

## RSM VS
RSM_2 <- read.csv("VS_RSMMLR-2.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
RSM_2 <- left_join(RSM_2, indices_2, by = "Index")

RSM_3 <- read.csv("VS_RSMMLR-3.csv", stringsAsFactors = FALSE, header = TRUE, sep = ",")
RSM_3 <- left_join(RSM_3, indices_3, by = "Index")

Concatenating datasets

## Concatenating datasets

set.seed(10)
Moleculas_2 <- as.data.frame(cbind(LASSO_2$Index, L_MLR_2$Index, GA_MLR_2$Index, 
    PLS_2$Index, RSM_2$Index))
colnames(Moleculas_2) <- c("LASSO", "LASSO-MLR", "GA-MLR", "PLS", "RSM")

Nombres_2 <- as.data.frame(cbind(LASSO_2$Label, L_MLR_2$Label, GA_MLR_2$Label, PLS_2$Label, 
    RSM_2$Label))
colnames(Nombres_2) <- c("LASSO", "LASSO-MLR", "GA-MLR", "PLS", "RSM")


Moleculas_3 <- as.data.frame(cbind(LASSO_3$Index, L_MLR_3$Index, GA_MLR_3$Index, 
    PLS_3$Index, RSM_3$Index))
colnames(Moleculas_3) <- c("LASSO", "LASSO-MLR", "GA-MLR", "PLS", "RSM")

Nombres_3 <- as.data.frame(cbind(LASSO_3$Label, L_MLR_3$Label, GA_MLR_3$Label, PLS_3$Label, 
    RSM_3$Label))
colnames(Nombres_3) <- c("LASSO", "LASSO-MLR", "GA-MLR", "PLS", "RSM")

Frequency of apparition of the molecules

## Two variants
Frecuencia_2 <- table(unlist(Moleculas_2[1:20, ]))
Frecuencia_2 <- as.data.frame(Frecuencia_2)
colnames(Frecuencia_2) <- c("Molecule", "Occurrence frequency 2")
Frecuencia_2$Molecule <- as.numeric(as.character(Frecuencia_2$Molecule))

## Three variants
Frecuencia_3 <- table(unlist(Moleculas_3[1:20, ]))
Frecuencia_3 <- as.data.frame(Frecuencia_3)
colnames(Frecuencia_3) <- c("Molecule", "Occurrence frequency 3")
Frecuencia_3$Molecule <- as.numeric(as.character(Frecuencia_3$Molecule))

Molecule frequency per QSAR model

Two variants

## Matching each frequency with initial index

I_INDEX_2 <- match(Frecuencia_2$Molecule, Moleculas_2$LASSO)
II_INDEX_2 <- match(Frecuencia_2$Molecule, Moleculas_2$`LASSO-MLR`)
III_INDEX_2 <- match(Frecuencia_2$Molecule, Moleculas_2$`GA-MLR`)
IV_INDEX_2 <- match(Frecuencia_2$Molecule, Moleculas_2$PLS)
V_INDEX_2 <- match(Frecuencia_2$Molecule, Moleculas_2$RSM)

Frecuencia_2_final <- as.data.frame(cbind(Frecuencia_2$Molecule, I_INDEX_2, II_INDEX_2, 
    III_INDEX_2, IV_INDEX_2, V_INDEX_2))
colnames(Frecuencia_2_final) <- c("Molecule", "LASSO", "LASSO_MLR", "GA_MLR", "PLS", 
    "RSM")
kable(Frecuencia_2_final[1:20, ], align = "cccccc")
Molecule LASSO LASSO_MLR GA_MLR PLS RSM
1 21 21 11 22 192
2 22 22 12 27 193
5 25 25 1 118 1
6 26 26 2 135 2
7 11 11 143 8 64
8 12 12 144 3 65
11 29 29 3 121 3
12 30 30 4 63 4
21 13 13 149 6 68
22 14 14 150 1 69
29 15 15 209 17 129
30 16 16 210 21 130
35 49 49 10 20 163
42 56 56 13 23 194
46 60 60 268 12 298
55 69 69 14 13 117
71 85 85 6 14 122
72 86 86 7 10 123
82 96 96 15 16 195
85 99 99 16 18 196

Three variants

I_INDEX_3 <- match(Frecuencia_3$Molecule, Moleculas_3$LASSO)
II_INDEX_3 <- match(Frecuencia_3$Molecule, Moleculas_3$`LASSO-MLR`)
III_INDEX_3 <- match(Frecuencia_3$Molecule, Moleculas_3$`GA-MLR`)
IV_INDEX_3 <- match(Frecuencia_3$Molecule, Moleculas_3$PLS)
V_INDEX_3 <- match(Frecuencia_3$Molecule, Moleculas_3$RSM)

Frecuencia_3_final <- as.data.frame(cbind(Frecuencia_3$Molecule, I_INDEX_3, II_INDEX_3, 
    III_INDEX_3, IV_INDEX_3, V_INDEX_3))
colnames(Frecuencia_3_final) <- c("Molecule", "LASSO", "LASSO_MLR", "GA_MLR", "PLS", 
    "RSM")
kable(Frecuencia_3_final[1:20, ], align = "cccccc")
Molecule LASSO LASSO_MLR GA_MLR PLS RSM
1 31 31 13 28 286
7 37 37 1 195 1
8 38 38 2 175 2
9 39 39 3 129 3
10 16 16 215 1 98
11 17 17 216 38 99
12 18 18 217 5 100
16 43 43 4 155 4
17 44 44 5 117 5
18 45 45 6 159 6
31 19 19 224 42 104
32 20 20 225 43 105
43 22 22 314 20 197
54 75 75 12 24 250
59 80 80 410 10 452
62 83 83 14 15 287
68 89 89 411 12 453
83 104 104 15 48 181
84 105 105 16 37 182
121 142 142 17 18 288

Top molecules

Top 20 molecules two variants

Moleculas_2_Total <- cbind(Moleculas_2, Nombres_2)
Moleculas_2_Total <- Moleculas_2_Total[, c(1, 6, 2, 7, 3, 8, 4, 9, 5, 10)]

kable(Moleculas_2_Total[1:20, ], align = "cccccccccc")
LASSO LASSO.1 LASSO-MLR LASSO-MLR.1 GA-MLR GA-MLR.1 PLS PLS.1 RSM RSM.1
115 446101 115 446101 5 118707521 22 54758613 5 118707521
116 446101 116 446101 6 118707521 125 440271 6 118707521
215 ZINC000019893747 215 ZINC000019893747 11 76325328 8 91820057 11 76325328
216 ZINC000019893747 216 ZINC000019893747 12 76325328 126 440271 12 76325328
275 ZINC000257346903 275 ZINC000257346903 246 ZINC000101361998 127 440235 147 198515
276 ZINC000257346903 276 ZINC000257346903 71 12600646 21 54758613 148 198515
277 ZINC000257347638 277 ZINC000257347638 72 12600646 128 440235 149 198514
278 ZINC000257347638 278 ZINC000257347638 96 5284365 7 91820057 277 ZINC000257347638
279 ZINC000257347639 279 ZINC000257347639 245 ZINC000101361998 191 739 278 ZINC000257347638
280 ZINC000257347639 280 ZINC000257347639 35 46174147 72 12600646 279 ZINC000257347639
7 91820057 7 91820057 1 134695375 96 5284365 280 ZINC000257347639
8 91820057 8 91820057 2 134695375 46 45266741 99 3052142
21 54758613 21 54758613 42 45357367 55 42612215 100 3052142
22 54758613 22 54758613 55 42612215 71 12600646 150 198514
29 51351654 29 51351654 82 7067772 86 6713972 219 ZINC000032106887
30 51351654 30 51351654 85 6713972 82 7067772 220 ZINC000032106887
125 440271 125 440271 86 6713972 29 51351654 221 ZINC000032106889
126 440271 126 440271 162 112106 85 6713972 222 ZINC000032106889
127 440235 127 440235 207 ZINC000015205840 230 ZINC000082114392 153 129618
128 440235 128 440235 208 ZINC000015205840 35 46174147 154 129618

Top 20 molecules three variants

Moleculas_3_Total <- cbind(Moleculas_3, Nombres_3)
Moleculas_3_Total <- Moleculas_3_Total[, c(1, 6, 2, 7, 3, 8, 4, 9, 5, 10)]

kable(Moleculas_3_Total[1:20, ], align = "cccccccccc")
LASSO LASSO.1 LASSO-MLR LASSO-MLR.1 GA-MLR GA-MLR.1 PLS PLS.1 RSM RSM.1
172 446101 172 446101 7 118707521 10 91820057 7 118707521
173 446101 173 446101 8 118707521 187 440271 8 118707521
174 446101 174 446101 9 118707521 192 440235 9 118707521
322 ZINC000019893747 322 ZINC000019893747 16 76325328 190 440235 16 76325328
323 ZINC000019893747 323 ZINC000019893747 17 76325328 12 91820057 17 76325328
324 ZINC000019893747 324 ZINC000019893747 18 76325328 172 446101 18 76325328
412 ZINC000257346903 412 ZINC000257346903 367 ZINC000101361998 189 440271 220 198515
413 ZINC000257346903 413 ZINC000257346903 369 ZINC000101361998 191 440235 221 198515
414 ZINC000257346903 414 ZINC000257346903 143 5284365 362 ZINC000101136391 222 198515
415 ZINC000257347638 415 ZINC000257347638 144 5284365 59 45933887 224 198514
416 ZINC000257347638 416 ZINC000257347638 368 ZINC000101361998 144 5284365 225 198514
417 ZINC000257347638 417 ZINC000257347638 54 46174147 68 45266741 415 ZINC000257347638
418 ZINC000257347639 418 ZINC000257347639 1 134695375 426 ZINC000307565252 416 ZINC000257347638
419 ZINC000257347639 419 ZINC000257347639 62 45357367 143 5284365 417 ZINC000257347638
420 ZINC000257347639 420 ZINC000257347639 83 42612215 62 45357367 418 ZINC000257347639
10 91820057 10 91820057 84 42612215 128 6713972 419 ZINC000257347639
11 91820057 11 91820057 121 7067772 129 6713972 420 ZINC000257347639
12 91820057 12 91820057 122 7067772 121 7067772 148 3052142
31 54758613 31 54758613 123 7067772 123 7067772 229 129618
32 54758613 32 54758613 127 6713972 43 51351654 231 129618

Applying an exponential consensus ranking (ECR) to the prediction to improve the selection of molecules

Recently, (Palacio-Rodríguez et al. 2019) proposed a new ranking function for virtual screening in molecular docking. According to the paper, this ranking function can be extrapolated into other ranking problems, such as this QSAR prediction. The function is:

\[\begin{align} \mathbf{P(i)} = \mathbf{\frac{1}{\sigma}\sum_{j}^{} \exp(-r_i^j)} \end{align}\]

Where i are the i-th molecule predicted with the ranking given by the j-th model (QSAR). \(\sigma\) is the expected value of the exponential distribution. This parameter takes account the number of molecules for each scoring function that will be considered (it can be seen as a threshold of the molecules that will be take account for the consensus). Finally, the sum gives the consensus ranking P(i) for the i-th molecule. Let’s apply this strategy

ECR function

ECR <- function(X, Sig) {
    
    EC <- (exp(-X/Sig))/Sig
    
    output <- EC
    
    return(output)
}

ECR for the QSAR models

## Two variants
ECR_2 <- cbind(match(indices_2$Index, Moleculas_2_Total$LASSO), match(indices_2$Index, 
    Moleculas_2_Total$`LASSO-MLR`), match(indices_2$Index, Moleculas_2_Total$`GA-MLR`), 
    match(indices_2$Index, Moleculas_2_Total$PLS), match(indices_2$Index, Moleculas_2_Total$RSM))
colnames(ECR_2) <- c("LASSO", "LASSO-MLR", "GA-MLR", "PLS", "RSM")

ECR_Salida_2 <- t(as.data.frame(apply(ECR_2, 1, ECR, Sig = 20)))

Ranking_Consenso_2 <- as.data.frame(apply(ECR_Salida_2, 1, sum))
colnames(Ranking_Consenso_2) <- c("ECR")
row.names(Ranking_Consenso_2) <- as.vector(indices_2$Index)
Ranking_Consenso_2$ID <- as.vector(indices_2$Index)

Output_ECR_2 <- Ranking_Consenso_2[with(Ranking_Consenso_2, order(Ranking_Consenso_2$ECR, 
    decreasing = TRUE)), ]

colnames(Output_ECR_2) <- c("ECR", "Index")
Output_ECR_2 <- left_join(Output_ECR_2, indices_2, by = "Index")
kable(Output_ECR_2[1:20, ], align = "ccc")
ECR Index Label
0.1239104 5 118707521
0.1177955 6 118707521
0.1096457 11 76325328
0.1063287 12 76325328
0.1039851 277 ZINC000257347638
0.0998926 8 91820057
0.0989136 278 ZINC000257347638
0.0988349 22 54758613
0.0987525 115 446101
0.0940895 279 ZINC000257347639
0.0932883 7 91820057
0.0925655 116 446101
0.0909432 21 54758613
0.0895007 280 ZINC000257347639
0.0889071 125 440271
0.0864304 215 ZINC000019893747
0.0824722 126 440271
0.0822214 216 ZINC000019893747
0.0804882 1 134695375
0.0784499 127 440235 
## Three variants
ECR_3 <- cbind(match(indices_3$Index, Moleculas_3_Total$LASSO), match(indices_3$Index, 
    Moleculas_3_Total$`LASSO-MLR`), match(indices_3$Index, Moleculas_3_Total$`GA-MLR`), 
    match(indices_3$Index, Moleculas_3_Total$PLS), match(indices_3$Index, Moleculas_3_Total$RSM))
colnames(ECR_3) <- c("LASSO", "LASSO-MLR", "GA-MLR", "PLS", "RSM")

ECR_Salida_3 <- t(as.data.frame(apply(ECR_3, 1, ECR, Sig = 20)))

Ranking_Consenso_3 <- as.data.frame(apply(ECR_Salida_3, 1, sum))
colnames(Ranking_Consenso_3) <- c("ECR")
row.names(Ranking_Consenso_3) <- as.vector(indices_3$Index)
Ranking_Consenso_3$ID <- as.vector(indices_3$Index)

Ranking_Consenso_3 <- as.data.frame(apply(ECR_Salida_3, 1, sum))
colnames(Ranking_Consenso_3) <- c("ECR")
row.names(Ranking_Consenso_3) <- as.vector(indices_3$Index)
Ranking_Consenso_3$ID <- as.vector(indices_3$Index)

Output_ECR_3 <- Ranking_Consenso_3[with(Ranking_Consenso_3, order(Ranking_Consenso_3$ECR, 
    decreasing = TRUE)), ]

colnames(Output_ECR_3) <- c("ECR", "Index")
Output_ECR_3 <- left_join(Output_ECR_3, indices_3, by = "Index")
kable(Output_ECR_3[1:20, ], align = "ccc")
ECR Index Label
0.1323101 172 446101
0.1108496 7 118707521
0.1054485 8 118707521
0.1003772 9 118707521
0.0945118 173 446101
0.0935430 16 76325328
0.0928678 10 91820057
0.0910086 322 ZINC000019893747
0.0891044 17 76325328
0.0880936 415 ZINC000257347638
0.0867441 174 446101
0.0846394 18 76325328
0.0837973 416 ZINC000257347638
0.0799349 12 91820057
0.0797105 417 ZINC000257347638
0.0779056 323 ZINC000019893747
0.0758229 418 ZINC000257347639
0.0744185 412 ZINC000257346903
0.0741063 324 ZINC000019893747
0.0740046 187 440271

Removing variants from the predictions (getting only the first variant per ligand according to each QSAR model)

Two variants

Moleculas_2_curado <- cbind(Moleculas_2_Total$LASSO[!duplicated(Moleculas_2_Total$LASSO.1)], 
    Moleculas_2_Total$LASSO.1[!duplicated(Moleculas_2_Total$LASSO.1)], Moleculas_2_Total$`LASSO-MLR`[!duplicated(Moleculas_2_Total$`LASSO-MLR.1`)], 
    Moleculas_2_Total$`LASSO-MLR.1`[!duplicated(Moleculas_2_Total$`LASSO-MLR.1`)], 
    Moleculas_2_Total$`GA-MLR`[!duplicated(Moleculas_2_Total$`GA-MLR.1`)], Moleculas_2_Total$`GA-MLR.1`[!duplicated(Moleculas_2_Total$`GA-MLR.1`)], 
    Moleculas_2_Total$PLS[!duplicated(Moleculas_2_Total$PLS.1)], Moleculas_2_Total$PLS.1[!duplicated(Moleculas_2_Total$PLS.1)], 
    Moleculas_2_Total$RSM[!duplicated(Moleculas_2_Total$RSM.1)], Moleculas_2_Total$RSM.1[!duplicated(Moleculas_2_Total$RSM.1)])

colnames(Moleculas_2_curado) <- c("Lasso_Index", "Lasso_Molecule", "Lasso-MLR_Index", 
    "Lasso-MLR_Molecule", "GA-MLR_Index", "GA-MLR_Molecule", "PLS_Index", "PLS_Molecule", 
    "RSM_Index", "RSM_Molecule")

Moleculas_2_curado[, 1] <- ceiling(as.numeric(Moleculas_2_curado[, 1])/2)
Moleculas_2_curado[, 3] <- ceiling(as.numeric(Moleculas_2_curado[, 3])/2)
Moleculas_2_curado[, 5] <- ceiling(as.numeric(Moleculas_2_curado[, 5])/2)
Moleculas_2_curado[, 7] <- ceiling(as.numeric(Moleculas_2_curado[, 7])/2)
Moleculas_2_curado[, 9] <- ceiling(as.numeric(Moleculas_2_curado[, 9])/2)

kable(Moleculas_2_curado[1:20, ], align = "cccccccccc")
Lasso_Index Lasso_Molecule Lasso-MLR_Index Lasso-MLR_Molecule GA-MLR_Index GA-MLR_Molecule PLS_Index PLS_Molecule RSM_Index RSM_Molecule
58 446101 58 446101 3 118707521 11 54758613 3 118707521
108 ZINC000019893747 108 ZINC000019893747 6 76325328 63 440271 6 76325328
138 ZINC000257346903 138 ZINC000257346903 123 ZINC000101361998 4 91820057 74 198515
139 ZINC000257347638 139 ZINC000257347638 36 12600646 64 440235 75 198514
140 ZINC000257347639 140 ZINC000257347639 48 5284365 96 739 139 ZINC000257347638
4 91820057 4 91820057 18 46174147 36 12600646 140 ZINC000257347639
11 54758613 11 54758613 1 134695375 48 5284365 50 3052142
15 51351654 15 51351654 21 45357367 23 45266741 110 ZINC000032106887
63 440271 63 440271 28 42612215 28 42612215 111 ZINC000032106889
64 440235 64 440235 41 7067772 43 6713972 77 129618
1 134695375 1 134695375 43 6713972 41 7067772 9 57339290
2 129878031 2 129878031 81 112106 15 51351654 142 ZINC000307565252
3 118707521 3 118707521 104 ZINC000015205840 115 ZINC000082114392 137 ZINC000257346492
5 90659182 5 90659182 105 ZINC000015205946 18 46174147 71 312829
6 76325328 6 76325328 132 ZINC000106384102 1 134695375 78 127132
7 57376616 7 57376616 32 24779679 21 45357367 83 99461
8 57339292 8 57339292 25 44341508 104 ZINC000015205840 118 ZINC000100815672
9 57339290 9 57339290 82 102416 9 57339290 17 49802606
10 54758653 10 54758653 85 90220 81 112106 130 ZINC000106382957
12 52948856 12 52948856 87 82398 142 ZINC000307565252 131 ZINC000106382965

Three variants

Moleculas_3_curado <- cbind(Moleculas_3_Total$LASSO[!duplicated(Moleculas_3_Total$LASSO.1)], 
    Moleculas_3_Total$LASSO.1[!duplicated(Moleculas_3_Total$LASSO.1)], Moleculas_3_Total$`LASSO-MLR`[!duplicated(Moleculas_3_Total$`LASSO-MLR.1`)], 
    Moleculas_3_Total$`LASSO-MLR.1`[!duplicated(Moleculas_3_Total$`LASSO-MLR.1`)], 
    Moleculas_3_Total$`GA-MLR`[!duplicated(Moleculas_3_Total$`GA-MLR.1`)], Moleculas_3_Total$`GA-MLR.1`[!duplicated(Moleculas_3_Total$`GA-MLR.1`)], 
    Moleculas_3_Total$PLS[!duplicated(Moleculas_3_Total$PLS.1)], Moleculas_3_Total$PLS.1[!duplicated(Moleculas_3_Total$PLS.1)], 
    Moleculas_3_Total$RSM[!duplicated(Moleculas_3_Total$RSM.1)], Moleculas_3_Total$RSM.1[!duplicated(Moleculas_3_Total$RSM.1)])

colnames(Moleculas_3_curado) <- c("Lasso_Index", "Lasso_Molecule", "Lasso-MLR_Index", 
    "Lasso-MLR_Molecule", "GA-MLR_Index", "GA-MLR_Molecule", "PLS_Index", "PLS_Molecule", 
    "RSM_Index", "RSM_Molecule")

Moleculas_3_curado[, 1] <- ceiling(as.numeric(Moleculas_3_curado[, 1])/3)
Moleculas_3_curado[, 3] <- ceiling(as.numeric(Moleculas_3_curado[, 3])/3)
Moleculas_3_curado[, 5] <- ceiling(as.numeric(Moleculas_3_curado[, 5])/3)
Moleculas_3_curado[, 7] <- ceiling(as.numeric(Moleculas_3_curado[, 7])/3)
Moleculas_3_curado[, 9] <- ceiling(as.numeric(Moleculas_3_curado[, 9])/3)

kable(Moleculas_3_curado[1:20, ], align = "cccccccccc")
Lasso_Index Lasso_Molecule Lasso-MLR_Index Lasso-MLR_Molecule GA-MLR_Index GA-MLR_Molecule PLS_Index PLS_Molecule RSM_Index RSM_Molecule
58 446101 58 446101 3 118707521 4 91820057 3 118707521
108 ZINC000019893747 108 ZINC000019893747 6 76325328 63 440271 6 76325328
138 ZINC000257346903 138 ZINC000257346903 123 ZINC000101361998 64 440235 74 198515
139 ZINC000257347638 139 ZINC000257347638 48 5284365 58 446101 75 198514
140 ZINC000257347639 140 ZINC000257347639 18 46174147 121 ZINC000101136391 139 ZINC000257347638
4 91820057 4 91820057 1 134695375 20 45933887 140 ZINC000257347639
11 54758613 11 54758613 21 45357367 48 5284365 50 3052142
15 51351654 15 51351654 28 42612215 23 45266741 77 129618
63 440271 63 440271 41 7067772 142 ZINC000307565252 110 ZINC000032106887
64 440235 64 440235 43 6713972 21 45357367 111 ZINC000032106889
1 134695375 1 134695375 44 6420074 43 6713972 9 57339290
2 129878031 2 129878031 53 2733335 41 7067772 108 ZINC000019893747
3 118707521 3 118707521 81 112106 15 51351654 142 ZINC000307565252
5 90659182 5 90659182 104 ZINC000015205840 44 6420074 137 ZINC000257346492
6 76325328 6 76325328 105 ZINC000015205946 18 46174147 71 312829
7 57376616 7 57376616 132 ZINC000106384102 53 2733335 78 127132
8 57339292 8 57339292 32 24779679 1 134695375 83 99461
9 57339290 9 57339290 25 44341508 104 ZINC000015205840 118 ZINC000100815672
10 54758653 10 54758653 82 102416 9 57339290 138 ZINC000257346903
12 52948856 12 52948856 85 90220 126 ZINC000102974493 13 52944904

ECR two variants

Output_ECR_2_curado <- cbind(Output_ECR_2$ECR[!duplicated(Output_ECR_2$Label)], Output_ECR_2$Index[!duplicated(Output_ECR_2$Label)], 
    Output_ECR_2$Label[!duplicated(Output_ECR_2$Label)])
colnames(Output_ECR_2_curado) <- c("ECR", "Index", "Molecule")

Output_ECR_2_curado[, 2] <- ceiling(as.numeric(Output_ECR_2_curado[, 2])/2)

kable(Output_ECR_2_curado[1:20, ], align = "ccc")
ECR Index Molecule
0.123910394377029 3 118707521
0.10964571955221 6 76325328
0.103985062328752 139 ZINC000257347638
0.0998926021126605 4 91820057
0.0988349376415874 11 54758613
0.0987524816241701 58 446101
0.0940895291726436 140 ZINC000257347639
0.0889070691073631 63 440271
0.0864304316353736 108 ZINC000019893747
0.0804882060518684 1 134695375
0.0784499443559222 64 440235
0.078082494327749 138 ZINC000257346903
0.0686878753936395 15 51351654
0.0670244674612893 36 12600646
0.0628736414964675 48 5284365
0.0573643004621222 18 46174147
0.0542501126233325 28 42612215
0.0493435166015162 9 57339290
0.0480181977442779 21 45357367
0.0469106652810004 41 7067772

ECR three variants

Output_ECR_3_curado <- cbind(Output_ECR_3$ECR[!duplicated(Output_ECR_3$Label)], Output_ECR_3$Index[!duplicated(Output_ECR_3$Label)], 
    Output_ECR_3$Label[!duplicated(Output_ECR_3$Label)])
colnames(Output_ECR_3_curado) <- c("ECR", "Index", "Molecule")

Output_ECR_3_curado[, 2] <- ceiling(as.numeric(Output_ECR_3_curado[, 2])/3)
kable(Output_ECR_3_curado[1:20, ], align = "ccc")
ECR Index Molecule
0.132310100606202 58 446101
0.110849573814621 3 118707521
0.0935430282121767 6 76325328
0.0928677690607199 4 91820057
0.0910086285115274 108 ZINC000019893747
0.0880936480494446 139 ZINC000257347638
0.0758229257026819 140 ZINC000257347639
0.0744184874508252 138 ZINC000257346903
0.0740045912479688 63 440271
0.0656928404795316 64 440235
0.0596569652183867 1 134695375
0.0592038919518806 48 5284365
0.0516837253833607 15 51351654
0.0500240637735259 21 45357367
0.0450734356889405 11 54758613
0.0448522533185709 18 46174147
0.0417817679463113 41 7067772
0.0400215026402701 43 6713972
0.0352496273855809 123 ZINC000101361998
0.0352353230855984 74 198515

Removing LASSO predictions since this QSAR and LASSO-MLR have the same output prediction and are biasing ECR algorithm

Two variants

ECR_2_final <- cbind(match(indices_2$Index, Moleculas_2_Total$`LASSO-MLR`), match(indices_2$Index, 
    Moleculas_2_Total$`GA-MLR`), match(indices_2$Index, Moleculas_2_Total$PLS), match(indices_2$Index, 
    Moleculas_2_Total$RSM))
colnames(ECR_2_final) <- c("LASSO-MLR", "GA-MLR", "PLS", "RSM")

ECR_Salida_2_final <- t(as.data.frame(apply(ECR_2_final, 1, ECR, Sig = 20)))

Ranking_Consenso_2_final <- as.data.frame(apply(ECR_Salida_2_final, 1, sum))
colnames(Ranking_Consenso_2_final) <- c("ECR")
row.names(Ranking_Consenso_2_final) <- as.vector(indices_2$Index)
Ranking_Consenso_2_final$ID <- as.vector(indices_2$Index)

Output_ECR_2_final <- Ranking_Consenso_2_final[with(Ranking_Consenso_2_final, order(Ranking_Consenso_2_final$ECR, 
    decreasing = TRUE)), ]

colnames(Output_ECR_2_final) <- c("ECR", "Index")
Output_ECR_2_final <- left_join(Output_ECR_2_final, indices_2, by = "Index")

Output_ECR_2_final <- cbind(Output_ECR_2_final$ECR[!duplicated(Output_ECR_2_final$Label)], 
    Output_ECR_2_final$Index[!duplicated(Output_ECR_2_final$Label)], Output_ECR_2_final$Label[!duplicated(Output_ECR_2_final$Label)])
colnames(Output_ECR_2_final) <- c("ECR", "Index", "Molecule")


Output_ECR_2_final[, 2] <- ceiling(as.numeric(Output_ECR_2_final[, 2])/2)

kable(Output_ECR_2_final[1:20, ], align = "ccc")
ECR Index Molecule
0.109585154534019 3 118707521
0.0979172051475202 6 76325328
0.074005672452017 11 54758613
0.0724520203079591 4 91820057
0.0687506578428166 139 ZINC000257347638
0.0675363225099268 63 440271
0.0663460395106793 36 12600646
0.0629913185963106 1 134695375
0.0626693029245443 48 5284365
0.0622081215915549 140 ZINC000257347639
0.0591128931831972 64 440235
0.0530496211371537 18 46174147
0.0526628308044291 28 42612215
0.0511910103991344 58 446101
0.0464991779285494 41 7067772
0.0453286089440921 43 6713972
0.0450695477565888 15 51351654
0.044977694613017 21 45357367
0.0433950328141207 108 ZINC000019893747
0.0406548194289939 9 57339290

Three variants

ECR_3_final <- cbind(match(indices_3$Index, Moleculas_3_Total$`LASSO-MLR`), match(indices_3$Index, 
    Moleculas_3_Total$`GA-MLR`), match(indices_3$Index, Moleculas_3_Total$PLS), match(indices_3$Index, 
    Moleculas_3_Total$RSM))
colnames(ECR_3_final) <- c("LASSO-MLR", "GA-MLR", "PLS", "RSM")

ECR_Salida_3_final <- t(as.data.frame(apply(ECR_3_final, 1, ECR, Sig = 20)))

Ranking_Consenso_3_final <- as.data.frame(apply(ECR_Salida_3_final, 1, sum))
colnames(Ranking_Consenso_3_final) <- c("ECR")
row.names(Ranking_Consenso_3_final) <- as.vector(indices_3$Index)
Ranking_Consenso_3_final$ID <- as.vector(indices_3$Index)

Output_ECR_3_final <- Ranking_Consenso_3_final[with(Ranking_Consenso_3_final, order(Ranking_Consenso_3_final$ECR, 
    decreasing = TRUE)), ]

colnames(Output_ECR_3_final) <- c("ECR", "Index")
Output_ECR_3_final <- left_join(Output_ECR_3_final, indices_3, by = "Index")

Output_ECR_3_final <- cbind(Output_ECR_3_final$ECR[!duplicated(Output_ECR_3_final$Label)], 
    Output_ECR_3_final$Index[!duplicated(Output_ECR_3_final$Label)], Output_ECR_3_final$Label[!duplicated(Output_ECR_3_final$Label)])
colnames(Output_ECR_3_final) <- c("ECR", "Index", "Molecule")

Output_ECR_3_final[, 2] <- ceiling(as.numeric(Output_ECR_3_final[, 2])/3)

kable(Output_ECR_3_final[1:20, ], align = "ccc")
ECR Index Molecule
0.102987715498939 3 118707521
0.0877188203235018 6 76325328
0.0847486293811661 58 446101
0.0704013208548588 4 91820057
0.0596793514049593 63 440271
0.0591908290240155 48 5284365
0.0577671150638129 139 ZINC000257347638
0.054279319942693 64 440235
0.0500720908576283 108 ZINC000019893747
0.0497206368646311 140 ZINC000257347639
0.0492358429492832 21 45357367
0.0490445665270496 1 134695375
0.0436763660257705 18 46174147
0.0417405127001481 41 7067772
0.0399924305596604 43 6713972
0.0391840829648896 138 ZINC000257346903
0.0352496270434173 123 ZINC000101361998
0.0352348647812116 74 198515
0.0350401711984567 15 51351654
0.031900896754871 121 ZINC000101136391

Top 3 molecules for each QSAR method

Two variants // LASSO & MLR-LASSO

Molecule 115

Molecule 115

Molecule 115_3D

Molecule 215

Molecule 215

Molecule 215_3D

Molecule 275

Molecule 275

Molecule 275_3D

Two variants // GA-MLR

Molecule 5

Molecule 5

Molecule 5_3D

Molecule 11

Molecule 11

Molecule 11_3D

Molecule 246

Molecule 246

Molecule 246_3D

Two variants // PLS

Molecule 22

Molecule 22

Molecule 22_3D

Molecule 125

Molecule 125

Molecule 125_3D

Molecule 8

Molecule 8

Molecule 8_3D

Two variants // RSM

Molecule 5

Molecule 5

Molecule 5_3D

Molecule 11

Molecule 11

Molecule 11_3D

Molecule 147

Molecule 147

Molecule 147_3D

Three variants // LASSO & MLR-LASSO

Molecule 172

Molecule 172

Molecule 172_3D

Molecule 322

Molecule 322

Molecule 322_3D

Molecule 412

Molecule 412

Molecule 412_3D

Three variants // GA-MLR

Molecule 7

Molecule 7

Molecule 7_3D

Molecule 16

Molecule 16

Molecule 16_3D

Molecule 367

Molecule 367

Molecule 367_3D

Three variants // PLS

Molecule 10

Molecule 10

Molecule 10_3D

Molecule 187

Molecule 187

Molecule 187_3D

Molecule 192

Molecule 192

Molecule 192_3D

Three variants // RSM

Molecule 7

Molecule 7

Molecule 7_3D

Molecule 16

Molecule 16

Molecule 16_3D

Molecule 220

Molecule 220

Molecule 220_3D

Top 3 molecules for exponential consensus ranking (ECR):

Two variants

Molecule 5

Molecule 5

Molecule 5_3D

Molecule 11

Molecule 11

Molecule 11_3D

Molecule 22

Molecule 22

Molecule 22_3D

Three variants

Molecule 7

Molecule 7

Molecule 7_3D

Molecule 16

Molecule 16

Molecule 16_3D

Molecule 123

Molecule 172

Molecule 172_3D

Molecular descriptors

One of the most important feature of these QSAR models are their physicochemical interpretation, since the correlation between these characteristics and the biological activity of the ligands can be used to propose new ligands and to get more insights about the mechanism of action.

Molecular descriptors

So, the common physicochemical features are the presence of C-S atoms at a certain topological distance, and at least one hydrogen bond donator and one hydrogen bond aceptor with a liphophilic group and at least one 6-membered ring. Please refer to the manuscript to get more information about

References

Palacio-Rodríguez, Karen, Isaias Lans, Claudio N. Cavasotto, and Pilar Cossio. 2019. “Exponential consensus ranking improves the outcome in docking and receptor ensemble docking.” Scientific Reports 9 (1): 1–14. https://doi.org/10.1038/s41598-019-41594-3.